Voice mode

The dashboard has a mic button — click it, talk, and the agent transcribes and acts on what you said, the same as if you'd typed it.

Privacy

Speech-to-text runs entirely on your own machine using a local model — your voice audio is never uploaded anywhere, including to us. This matters if you're using it during client calls or meetings where anything you say might be sensitive.

Two ways to use it

• Command mode — say something like "create a task to fix my car tomorrow" and it's acted on directly.
• Meeting mode — record a longer stretch of conversation (a meeting, a call) and get back a summary plus any action items it caught, including anything you explicitly flagged with a phrase like "make that a task."

If transcription seems off

Accuracy depends on the transcription model size — the default balances speed and accuracy well for most speech, but if it's consistently mishearing names or technical terms, a larger local model (configurable in Settings) trades a bit of speed for better accuracy.